lets make sure this can learn via supervision first

model.py

@@ -50,16 +50,24 @@ class ColorTransformerModel(L.LightningModule):
             outputs,
         )
         # alpha = self.hparams.alpha  # TODO: decide what to do with this...
-        loss = p_loss
+        # loss = p_loss
+
+        # distance = torch.minimum(
+        #     torch.norm(outputs - labels), torch.norm(1 + outputs - labels)
+        # ).mean()
+        distance = torch.norm(outputs - labels).mean()
+        loss = distance
+
+        self.log("train_loss", distance)
         self.log("hp_metric", loss)
         self.log("p_loss", p_loss)
-        return loss
+        return distance
 
     def validation_step(self, batch):
         inputs, labels = batch  # these are true HSV labels - no learning allowed.
         outputs = self.forward(inputs)
         distance = torch.minimum(
-            torch.abs(outputs - labels), torch.abs(1 + outputs - labels)
+            torch.norm(outputs - labels), torch.norm(1 + outputs - labels)
         )
         mean_loss = torch.mean(distance)
         max_loss = torch.max(distance)

newsearch.py

@@ -6,7 +6,7 @@ import numpy as np  # noqa: F401
 from lightning_sdk import Machine, Studio  # noqa: F401
 
 # consistency of randomly sampled experiments.
-seed(19920921)
+seed(202419920921)
 
 NUM_JOBS = 100
 
@@ -23,7 +23,7 @@ NUM_JOBS = 100
 # Define the ranges or sets of values for each hyperparameter
 # alpha_values = list(np.round(np.linspace(2, 4, 21), 4))
 # learning_rate_values = list(np.round(np.logspace(-5, -3, 21), 5))
-learning_rate_values = [1e-2]
+learning_rate_values = [1e-3]
 alpha_values = [0]
 widths = [2**k for k in range(4, 13)]
 depths = [1, 2, 4, 8, 16]
@@ -65,8 +65,8 @@ for idx, params in enumerate(search_params):
 python newmain.py fit \
 --seed_everything {s} \
 --data.batch_size {bs} \
---data.train_size 0 \
+--data.train_size 10000 \
---data.val_size 100000 \
+--data.val_size 10000 \
 --model.alpha {a} \
 --model.width {w} \
 --model.depth {d} \